omgwtfbyobbq
Active Member
For anyone who has purchased FSD, is there any actual contract that's different from Tesla's description? If so, that would the last word on what Tesla is legally required to provide.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Could the Model 3 autopilot HW support up to Level 5 autonomous driving?
- Thread starter coupedncal
- Start date
MotoEvCA
Member
While being a huge fan of Elon Musk, I do think it is time for Tesla to show a bit more restraint and not oversell (verbally) what they intend to offer in the future and publish (electronically) documentation in parallel with product releases / introductions. There is too much that requires speculation when dealing with Tesla features and products these days.
MotoEvCA
Member
I spoke with a Tesla agent regarding FSD and they acknowledged there is no contractual obligation to provide SPECIFIC FUNCTIONALITY THAT CORRESPONDS TO SAE AUTONOLOMY LEVELS. Nor is there a contractual obligation to refund if those autonomy levels are not achieved by Tesla as it relates to FSD.
I also asked if Tesla would be open to adding an addendum to the purchase contract regard FSD if included in the purchase contract.
Bottomline: The devil is in the details...
If Humans can drive in those conditions with literally only 2 eyes. Then a car with 8 eyes seeing at the same time, sensors and radar could too.
diplomat33
Average guy who loves autonomous vehicles
Picture time
Can human eyes handle those situations? The answer is yes because human drivers do handle them all the time. So if human eyes are good enough with the right software (brain) then surely high def cameras that are better than human eyes, with radar and ultrasonics that humans don't have, should be able to handle it with the right software too!
diplomat33
Average guy who loves autonomous vehicles
Perhaps, we are getting too caught up with the technical definitions of the SAE levels of autonomy? I mean, if the average Joe can get into a Tesla and it drives itself through traffic from their home to their workplace, they will consider the car to be self-driving and will be excited. They probably won't care if the self-driving was technically only level 4, not 5.
And as others have already stated, Tesla only promised self-driving in "most instances" which is not L5 autonomy. So, Tesla just needs to deliver a car that can drive itself in the typical commute and on highways, to meet its promises and satisfy most customers. It does not need to achieve L5 autonomy right away.
I do think we will see L3 and L4 autonomy first, before we see L5 autonomy. In that regards, I do expect Tesla to achieve L3 autonomy next year.
And as others have already stated, Tesla only promised self-driving in "most instances" which is not L5 autonomy. So, Tesla just needs to deliver a car that can drive itself in the typical commute and on highways, to meet its promises and satisfy most customers. It does not need to achieve L5 autonomy right away.
I do think we will see L3 and L4 autonomy first, before we see L5 autonomy. In that regards, I do expect Tesla to achieve L3 autonomy next year.
I agree, we should not get caught up in definitions. However, it is easier to talk about something if there are definitions like what each level of autonomy is.
Bottom line, they promised Tesla network (pre-req is fsd) of taxis without anyone inside. Thus is effectively not lvl 1,2, or 3. Only level 4 and 5 can be what they promised and level 5 does not require a steering wheel I think will be needed as they said they're covering almost everything.
Here is where human eyes can drive vs fsd lvl 4 can't.
Attachments
I'm *pretty sure* a human uses a bit more than just their eyes for figuring out how to drive in complex situations. I think I see what you're getting at though.. and that this is not a hardware problem, it's a software problem.
But a human is a lot more intelligent, than a car. The real question is how clever does a computer need to be to interpret the information from just vision and radar (and a bit of sonar).
lunitiks
Cool James & Black Teacher
Sure, forget blind spots, clogged up or iced down cameras and/or radar Heck, why do we even need sensors in the first place -- with clever enough algorithms I'm sure the car would be able to drive without any sensor input. It's just a software problem!
(Btw @diplomat33 I replied to your statement that the current sensor HW is "enough for L5 autonomy". )
Heck, why do we even need sensors in the first place -- with clever enough algorithms I'm sure the car would be able to drive without any sensor input. It's just a software problem!(Btw @diplomat33 I replied to your statement that the current sensor HW is "enough for L5 autonomy". )
Sure, forget blind spots, clogged up or iced down cameras and/or radar
(Btw @diplomat33 I replied to your statement that the current sensor HW is "enough for L5 autonomy". )
Not quite sure what you're getting at here, but it *is* a software problem. We (humans) when driving a car have an almost immeasurable amount of machine learning packed into our brain. That is the reason why we can navigate such complex driving situations with relative ease. We, as our friendly cars, are indeed limited to what we can see and feel with our eyes and our sensory systems. The hardware that is in today's cars can *in theory* replicate our senses, what can't be replicated is our brain. We are a *long* way off from that (IMO).
diplomat33
Average guy who loves autonomous vehicles
Sure, forget blind spots, clogged up or iced down cameras and/or radar
Clearly, you are being sarcastic. With enough cameras, the car would not have blind spots so that would not be an issue. And you could easily use heated elements to keep ice off the cameras and radar. Or the driver could remove the ice from their car like they normally do anyway. So I fail to see the problem. FSD is a software issue because regardless of what sensors you use, the car still needs a "brain" to process the data in order to actually make the right driving decisions. Even with LIDAR, the car still needs software to make FSD work. LIDAR is not some magic FSD system. What if the LIDAR sensor gets ice on it or hail hits it and damages it? It's easy to think of situations that might impair a FSD system. That does not mean that FSD is impossible.
How about we aim for what I call "partial FSD" before we rule out FSD completely just because we can imagine an extreme scenario where it might not work? Baby steps. Let's do L3 and L4 first. Then we can improve the system until it can drive in an extreme blizzard perfectly.
Yes, I saw it. And I replied again. See my post #26. I fail to see how the current sensors of 8 cameras, radar and ultrasonics would not be able to handle the scenarios you present, with the right software.
diplomat33
Average guy who loves autonomous vehicles
Perhaps, the issue is that people lack faith, so to speak. They see how easy it is for a human to drive in a snow storm and just think "there is no way a computer could handle that", so they conclude that FSD is impossible. But we humans have overcome some remarkable engineering obstacles that appeared impossible at first. FSD is definitely a big engineering challenge but I believe humans can figure out a way to make it happen!
lunitiks
Cool James & Black Teacher
That’s a simplistic and somewhat naive assumption. Humans have the advantage of an incredibly powerful nn, trained with a lifetime of contextual experience. We are a LONG way from the time when we have a computer sitting a Tesla that can equal it.
Automated driving may well be possible, but it will require far more that just two cameras. It will likely require car to car communication, hi res GPS, and fixed infrastructure upgrades.
diplomat33
Average guy who loves autonomous vehicles
I don't think I am being too naive. A sufficiently powerful computer with machine learning should be able to eventually approximate how a human drives.
Absolutely. That is why Tesla cars have 8 cameras! My point was not that since humans can drive with 2 eyes that FSD cars only need 2 cameras. My point was that if humans can drive with 2 cameras and a brain then surely a car with 8 cameras, radar, ultrasonics and a super computer, can drive itself too.
I have no expectation of level 5 with the model 3. I'll be buying FSD, and would be happy with level 4 autonomy in the most typical driving domains and think the investment would be worth every penny as an early adopter.
WileyTheMan
Peanut Gallery Member
The tech is in its infancy. Tesla is committed to bringing self-driving cars to market, and is equipping their cars the best they can for a future that is not clear. I will not throw a tantrum if the Model 3's equipment isn't up-to-snuff for Level 5 because of this. But I am also not purchasing FSD or EAP software updates until this future is a little more focused.
Yes, it could. Or maybe not. Anyone on this board who has a definitive answer is talking out their a**.
Similar threads
